---
name: academic-writing-structure-and-argumentation
description: >
  Builds argument into academic writing at paragraph, section and chapter level:
  claim, evidence, warrant and link; arguing with the literature rather than
  describing it; integrating sources into sentences; appropriate hedging and
  academic register; disagreeing with a published author; and editing for argument.
  Use for "my literature review is just a list", "how do I be more critical",
  "my writing is too descriptive", "how do I structure a paragraph", "feedback
  says needs more analysis", "how do I disagree with an author", "how do I write
  more academically", "my discussion just repeats my results".
category: 15 Academic University Research
ref: "15.05"
tier: 3
inherits: [K2, K3, K4, K5]
---

# Academic Writing Structure and Argumentation

## 1. One-line description

A method for making academic writing argue rather than describe: a paragraph structure that carries a claim and earns it, chapter architecture where each section makes the next necessary, source integration that supports a point instead of replacing it, calibrated hedging, and an editing pass that works on argument rather than on words.

## 2. What this skill is used for

**The research problem it solves.** The single largest difference between a pass and a distinction at undergraduate and honours level is not effort, reading or writing quality in the ordinary sense. It is whether the writing argues. Descriptive writing tells the reader what sources said, what was done and what was found. Analytical writing tells the reader what it means, why it matters, and why they should accept it. Most students can recognise the difference in someone else's work and cannot produce it in their own, because nobody has told them what the unit of production is. The unit is the paragraph, and a paragraph that argues has a definite structure: a claim, the evidence for it, the reasoning that makes the evidence count as support, and a link that makes the next claim necessary. Once that structure is visible, the classic failures become visible with it: the literature review that is a list of studies, the methodology that describes without justifying, the discussion that repeats the results, and the essay in which every paragraph is competent and the whole says nothing.

**Where it sits in the research lifecycle.** Throughout. It applies to the proposal, to every chapter of a dissertation, and to the final editing pass, and it is the skill most directly reflected in the marks for critical analysis, argument and structure.

**Typical use cases.**
- Feedback saying "too descriptive", "needs more critical analysis", or "you describe the literature but do not engage with it".
- A literature review that reads as a sequence of study summaries.
- A methodology chapter that says what was done and not why.
- A discussion chapter that restates the results chapter in different words.
- Learning how to disagree with a published author without either deference or rudeness.
- Getting hedging right, which students both overdo and underdo, sometimes in the same paragraph.
- Editing a complete draft for argument rather than for wording.

**Who uses it.** Undergraduate and honours students, especially those receiving consistent feedback about description and criticality that they do not know how to act on; students writing in a second language, for whom register and hedging are additional and separate problems; supervisors who need to explain the descriptive-to-analytical shift concretely rather than by exhortation.

## 3. When to use it

- Your feedback repeatedly says descriptive, uncritical, or lacking analysis, and you do not know what to change.
- Your literature review is organised as one paragraph per study.
- You can describe what each source says and cannot say what they collectively establish.
- Your methodology chapter would be equally true of somebody else's project.
- Your discussion chapter and your results chapter say the same things.
- You want to disagree with a published author and do not know whether you are allowed to, or how.
- Your writing feels either too casual or so formal that you cannot follow your own sentences.
- You have a complete draft and an editing week, and want to spend it on something that changes the mark.
- Your paragraphs are individually fine and the chapter does not add up.

## 4. When NOT to use it

- **The analysis has not been done.** No structural method converts absent findings into an argument. If the results are not yet analysed, or the sources are not yet read, writing is premature and rewriting will not help. Do the analysis, or read the sources, then return. Writing your way around missing evidence is the most expensive detour available, and it produces exactly the fluent, empty prose markers penalise.
- **The problem is what to cite or how.** Reference mechanics, paraphrase safety and citation verification are **15.03 Referencing and Citation Management**. This skill assumes verified sources and works on how they enter a sentence.
- **The problem is that you have not found enough literature.** A review cannot argue with a body of work it does not have. Go to **15.02 Academic Literature Search Strategy**.
- **The document is a proposal.** Proposals have their own assessed structure and their own alignment requirements, and the argument runs along a different thread. Use **15.04 Research Proposal Writing**, and apply this skill within its sections.
- **The task is proofreading, language correction or translation.** Grammar, spelling and sentence-level fluency are a different service, and in many institutions third-party language editing has its own declaration rules. This skill works on argument, not on correctness, and a well-argued draft with grammatical errors scores better than a flawless one that describes.
- **The discipline requires a fixed non-argumentative format.** Some laboratory reports, technical specifications, clinical documentation and standardised protocol write-ups are assessed on conformity to a template and completeness rather than on argument. Applying an argumentative structure to them is a mistake. Ask what the format demands before restructuring anything.
- **The output is a commercial or client-facing research report.** The architecture there is decision-led rather than argument-led, conclusions lead rather than follow, and the reader is acting rather than assessing. Use **12.03 Research Report Compilation**. Importing academic structure into a client report buries the answer.
- **The request is for passages to be written for submission.** See §12.1. This skill diagnoses, structures, critiques and teaches. It does not produce prose to be handed in, and the writing is the thing being assessed.

## 5. Required inputs

**Required.**
- **The draft, or the section being worked on.** This skill operates on text. Advice given without text is generic and does not transfer.
- **The controlling claim of the piece**, or an acknowledgement that there is not one yet, which is itself the finding that determines the work.
- **The document type and its assessed criteria.** What counts as good structure differs between a literature review chapter, a methodology chapter and a discussion, and the criteria say which qualities are being marked.

**Optional, and what each one adds.**
- **Previous feedback on the student's writing.** Recurring comments identify the specific habit to target, and one targeted habit change is worth more than a general improvement effort.
- **The chapter's position in the whole document.** Determines what the section must have established by its end for the next one to work, which is what section-level argument means.
- **A piece of writing in the discipline that the student admires.** The fastest route to register calibration, because academic register is field-specific and learned by pattern rather than by rule.
- **Whether English or the language of submission is the student's first language.** Changes the balance of the work: register and hedging need explicit treatment, while argument structure is language-independent and often already present.
- **The word count and how much is already used.** Determines whether the task is adding argument or cutting description to make room for it, which are different edits.
- **The evidence available**, for a discussion chapter. You cannot build an argument beyond what the results support, and the ceiling on the argument is set by the analysis (K3 §3.6).

## 6. Questions to ask before starting

1. **What is this document, and what are its assessed criteria?** A literature review, a methodology chapter and a discussion have different jobs and different failure modes. Default: ask; if unavailable, assume critical engagement and argument carry substantial weight, which is true in most rubrics.
2. **What is the one thing this chapter is arguing?** If the student cannot say it in a sentence, that is the diagnosis and the first task. Default: derive a candidate from the draft and test it against the text, showing the student where the text does and does not support it.
3. **Has the analysis or the reading actually been done?** Determines whether this is a writing task or a research task wearing a writing task's clothes. Default: check the draft for evidence rather than assertion before proceeding.
4. **What discipline, and what register does it use?** Formality, use of the first person, hedging conventions and how directly one may criticise vary substantially by field, and there is no universal academic style. Default: ask, and where you cannot, work from a sample of the field's writing rather than from a general rule.
5. **What is the recurring feedback?** Targets the work. Default: run the diagnostic in Step 1 and let the draft identify its own pattern.
6. **How much time and word count remain?** Determines whether to restructure or to intervene at paragraph level. Default: assume limited time, and prioritise the reverse outline and the first-sentence test, which give the most improvement per hour.
7. **What is your institution's AI use policy for this assessment, and what must be declared?** Governs how this skill may be used, and it matters more here than anywhere else in this track, because the writing is the thing assessed. Default: assume assistance must be declared and no generated text may be submitted (§12.1).

## 7. Step-by-step methodology

**Step 1. Diagnose description against argument, on the student's own text.**
Take two pages and mark every sentence as one of three types. **Reporting**: what a source said, what was done, what was found. **Analysis**: what it means, how it compares, why it differs, what follows. **Positioning**: the student's own claim about the matter. A descriptive draft is almost entirely reporting, with a positioning sentence at the end of each section that the reporting has not earned. A strong draft carries a positioning sentence early in each paragraph, reporting as its support, and analysis joining them. Count the proportions, because students consistently believe they are analysing when they are reporting, and a marked-up page ends that disagreement in thirty seconds. *Correct result: two pages colour-coded by sentence type, and a ratio the student can see.*

**Step 2. Rebuild the paragraph as a unit of argument.**
This is the highest-leverage thing an undergraduate can learn, and it is four moves.
- **Claim.** The paragraph's own point, stated first, in the student's voice. Not the topic of the paragraph but the assertion it will defend. "Several studies have examined switching behaviour" is a topic. "Studies of switching behaviour have measured intention rather than action, which limits what they establish about actual switching" is a claim.
- **Evidence.** The sources, data or findings that support it, with citations and locators.
- **Warrant.** The reasoning that makes this evidence count as support for this claim. This is the move students omit, and its absence is the technical reason a paragraph reads as descriptive: claim and evidence are both present, and the connection between them is left for the reader to construct. The warrant is usually one or two sentences stating the effect of what has just been shown, not a summary of it.
- **Link.** The sentence that makes the next paragraph necessary, either by opening the question the next claim answers or by naming the limitation the next paragraph addresses.
Test any paragraph by deleting everything except its first and last sentences. If those two do not carry a claim and a consequence, the paragraph is a container of material rather than a step in an argument. Two habits follow: one claim per paragraph, and the claim in the student's own words rather than a source's. *Correct result: rebuilt paragraphs where the first sentence states a claim, the middle earns it, and the last sentence obliges the next paragraph to exist.*

**Step 3. Build the section so each part earns the next.**
Write the first sentence of each paragraph in a section on one page, in order, and read only those. This reverse outline is the fastest structural diagnostic available and takes about ten minutes for a chapter. Three faults show immediately. If the sentences are topics rather than claims, the section is a container. If the claims could be reordered without loss, there is no argument, only a list. If a claim depends on something established two paragraphs later, the order is wrong. A section that argues reads, in its first sentences alone, as a chain: this is so, which means this, which is limited in this way, which is why this question arises. Then check the section against its job: what must a reader accept by its end for the next section to work? If nothing, the section is decoration. *Correct result: a first-sentence outline that reads as a connected argument, and a stated job for each section.*

**Step 4. Argue with the literature instead of describing it.**
The literature review that lists is the commonest structural failure in student work, and the fix is structural rather than stylistic: it is organised by source, and it must be organised by claim, theme or position. Four moves convert description into engagement, and a strong review uses all four.
- **Group.** Put sources together by what they claim, find or assume, so the paragraph's subject is a position rather than an author: "studies that measure intention" rather than "Author A, then Author B".
- **Compare.** State what differs between them and, crucially, why: different populations, definitions, measures, periods or designs. The reason for a difference is more valuable than the difference.
- **Evaluate.** Assess what the design can and cannot establish, what the sample supports, whether the conclusion exceeds the data. This is where criticality lives, and it requires having read the method sections rather than the abstracts.
- **Position.** Say where you stand and why, which is what makes it your review rather than a summary of the field.
The reliable signal of a list is that each paragraph begins with an author's name. The signal of engagement is that each begins with a claim about the literature and cites several sources inside it. *Correct result: a review organised by position, where most paragraphs cite more than one source and no paragraph is a single study's summary.*

**Step 5. Integrate sources so they support the point rather than replace it.**
Two forms, chosen deliberately. **Information-prominent** citation puts the claim first and the source in parentheses, and is the default, because it keeps your claim in the subject position: the finding is being used. **Author-prominent** citation puts the author in the sentence, and is used when the author is the subject: when you are comparing their position with another's, when they are being disputed, or when their authority is itself relevant. Students overuse the author-prominent form, which produces "Author A found... Author B found..." and puts the sources in charge of the paragraph. Two further rules. Introduce a source with a verb that does analytical work: found, argued, demonstrated, assumed, conceded and overlooked all say different things, while "talks about" and "states" say nothing. And never let a quotation carry the claim; the claim is yours and the quotation supports it, which means no paragraph ends on a quotation. *Correct result: paragraphs where the student's claim occupies the subject position and sources appear as support, with reporting verbs chosen for meaning.*

**Step 6. Take a critical stance, and disagree properly.**
Criticality is not negativity, and undergraduates commonly believe critical means finding fault. It means assessing what evidence supports, which frequently means agreeing with reasons. Where you do disagree with a published author, four rules keep it defensible. Represent their position accurately and at its strongest before responding, because disagreeing with a weakened version is visible and costs credibility. Aim at a specific claim rather than at the author or the paper as a whole. Give a reason of a recognised kind: the design cannot support the inference, the sample does not extend to the population claimed, the construct is measured differently from how it is defined, a plausible alternative explanation is unaddressed, or subsequent evidence points the other way. And calibrate the strength of the disagreement to the strength of your grounds. Note that a limitation the author has themselves acknowledged is weak criticism, and that the strongest criticism available at undergraduate level is usually a limit on transferability: what the study establishes, and where it stops establishing it. *Correct result: disagreements that name a specific claim, a recognised ground, and a proportionate strength.*

**Step 7. Calibrate hedging, and separate formality from obscurity.**
Academic writing requires hedging because claims exceed evidence by default, and students err in both directions, often within one paragraph: a strong finding hedged into invisibility, a speculative one asserted flatly. The rule is proportionality and it maps onto K3. A well-evidenced finding is stated plainly, because hedging a strong claim is miscommunication rather than modesty. A supported but contestable reading takes one hedge and its alternative explanation in the same passage. A speculation is labelled as one, with what would test it. One hedge is enough: "may possibly suggest that it could be the case that" is four hedges saying nothing, and a recognisable marker of defensive writing. Separately, register is not obscurity. Formal academic writing avoids contractions and colloquialism, prefers precision to emphasis, and does not use intensifiers to do the work of evidence. It does not require long words where short ones are exact, nominalisations that hide who did what, or sentences whose grammar has to be reconstructed. If you cannot follow your own sentence on a second reading, the sentence is failing, and complexity of expression is not evidence of complexity of thought. *Correct result: hedge strength matched to evidence strength throughout, and no sentence requiring a second pass to parse.*

**Step 8. Signpost enough to orient, and no more.**
Readers need to know where they are, which is what signposting is for. Mechanical signposting, where every section opens by announcing itself and closes by summarising itself, wastes words and reads as filler. Three techniques do the work more economically. Use the claim itself as the signpost: a first sentence stating the paragraph's point tells the reader where they are without a separate announcement. Use transitions that name a logical relation, so that "however", "consequently" and "by contrast" carry meaning, rather than sequence markers like "firstly", which carry none. And reserve explicit orientation for the joints that matter: the start of a chapter, and a genuine change of direction. *Correct result: a reader who can always say what the current claim is, achieved with a minimum of announcing.*

**Step 9. Fix the three chapter-level failures, each of which has a specific cause.**
- **The literature review that is a list.** Cause: organised by source. Fix: reorganise by claim or theme, per Step 4, which is a restructure rather than a rewrite and takes an afternoon with a card per source.
- **The methodology that describes without justifying.** Test: would this chapter be equally true of a different project on a different question? If yes, it is a textbook account. Fix: after every methodological statement, add the sentence beginning with why this suited this question, and name at least one rejected alternative. Justification requires a comparison; a decision with no alternative stated has not been justified.
- **The discussion that repeats the results.** Cause: no distinction held between finding and interpretation. Fix: enforce the K2 §2.1 boundary. Results state what was observed. Discussion states what it means, how it relates to the literature reviewed, where it agrees and disagrees with prior work and why, what alternatives exist, and what follows. A discussion paragraph with no citation and no interpretive move is a results paragraph in the wrong chapter. The most useful discussion paragraph structure is: finding, relation to prior work, interpretation, alternative explanation, limit.
*Correct result: each failure diagnosed by its own test and fixed at its cause rather than by adding words.*

**Step 10. Edit for argument before editing for words.**
Two passes, in this order, because the second is wasted if the first has not run. **Argument pass**: reverse outline the whole document from first sentences; check that each chapter has one controlling claim; check that every claim is supported and every piece of support is used for a claim; delete paragraphs that report without arguing; and apply the so-what test to each section, asking what a reader who accepted it would now have to accept. **Sentence pass**: only then work on wording, hedging, reporting verbs, transitions and register. Students reliably do these in reverse order and spend an editing week polishing paragraphs that will be cut. Where the word count is over, cut description rather than argument: reporting sentences that summarise a source without using it are the cheapest words in the document. *Correct result: a document whose first sentences alone read as a coherent argument, edited for language only after that is true.* `RESEARCHER REVIEW RECOMMENDED` on the controlling claim of the whole document, which depends on judgements about the field and the assessment that a supervisor holds (K5 §2.1).

## 8. Analytical framework

**The paragraph unit**, which is where argument is actually built:

    Claim → Evidence → Warrant → Link

Applying it: the warrant is the diagnostic. Nearly every descriptive paragraph in student work has a claim somewhere and evidence somewhere, and no warrant, which is why the reader is left to assemble the argument themselves and why the writing feels like a summary. Test by asking, of any evidence sentence, "and this shows what?" If the text does not answer, the warrant is missing. The link is the second most commonly absent move, and its absence is what produces a chapter of individually competent paragraphs that does not accumulate.

**The engagement ladder**, which is what "critical" means operationally:

    Report (X said this)
      → Group (these sources share this position)
        → Compare (they differ here, and this is why)
          → Evaluate (this evidence supports this much, and no further)
            → Position (therefore I hold this, on these grounds)

Applying it: a pass-level review sits on the first rung, and a distinction-level review moves up and down the ladder deliberately, reporting briefly where the material is settled and climbing to evaluation and position where the argument turns. The ladder also calibrates effort: not every paragraph needs to reach the top, but a chapter in which nothing does has not argued. Marking criteria that use the words "critical", "analytical" and "evaluative" are describing the upper three rungs.

## 9. Output format

This skill produces diagnosis and restructuring, not submittable prose. Its outputs are working artefacts.

**1. Sentence-type diagnostic.** A marked sample showing reporting, analysis and positioning sentences, with the ratio stated.

**2. Reverse outline.** The first sentence of every paragraph, in order, with each marked as a claim or a topic, and structural faults noted (unordered, dependent on later material, decorative).

**3. Controlling claim register.** One sentence per chapter, plus one for the document, with a note where the text does not support the claim stated.

**4. Paragraph audit.** For selected paragraphs: claim present, evidence present, warrant present, link present, and what is missing.

**5. Source integration review.** Instances of author-prominent citation where information-prominent would be stronger, reporting verbs doing no analytical work, quotations carrying claims, and paragraphs built on a single source.

**6. Hedging calibration.** Passages where the hedge does not match the evidence, in both directions, with the K3 level each should carry.

**7. Chapter-failure diagnosis.** Which of the three classic failures is present, its test, and the structural fix.

**8. Prioritised edit plan.** The argument-pass actions first, sentence-pass actions second, with an estimate of the word budget released by cutting description.

**When the underlying work is thin, the format must not force fabrication (K4 §1).** Where a paragraph has a claim and no evidence, the correct output is the observation that the claim is unsupported and the question of what evidence exists, never a suggested sentence that supplies plausible support. Where a discussion has no findings to interpret, say that the analysis is incomplete. Structural improvement cannot manufacture content, and a well-structured empty argument is easier for a marker to see through than a disorganised one.

## 10. Quality checks

Run on a complete draft. These sit on top of K4 §8.

1. Can you state the controlling claim of each chapter in one sentence, and does the chapter support it?
2. Read only the first sentences of each paragraph: do they form a connected argument, or a list of topics?
3. Does each paragraph make one claim, in your words, in its first sentence?
4. Does each paragraph contain a warrant, so the reader is not left to connect evidence to claim?
5. Does each paragraph's last sentence make the next paragraph necessary?
6. Could the paragraphs in any section be reordered without loss? If so, there is no argument in that section.
7. Does any paragraph begin with an author's name, and would a claim serve better?
8. Do most literature paragraphs cite more than one source?
9. Does any quotation carry a claim that should be yours, or end a paragraph?
10. Does every reporting verb do analytical work, or are they interchangeable?
11. Does the strength of every hedge match the strength of the evidence behind it (K3 §4)?
12. Is every disagreement with an author aimed at a specific claim, on a recognised ground, at proportionate strength?
13. Would the methodology chapter be equally true of a different project?
14. Does the discussion interpret, relate and qualify, or does it restate the results?
15. Is the boundary between finding and interpretation visible throughout (K2 §2.1)?
16. If over the word count, has description been cut before argument?

## 11. Common failure modes

| Failure | How to recognise it | How to prevent it |
|---|---|---|
| **Description mistaken for analysis** | Almost every sentence reports; the student believes they analysed | Sentence-type diagnostic on their own text (Step 1) |
| **The missing warrant** | Claim and evidence present, connection left to the reader | Ask "and this shows what?" of every evidence sentence (Step 2) |
| **Topic sentences instead of claim sentences** | First sentences announce subjects, not assertions | Reverse outline and rewrite first sentences as claims (Step 3) |
| **The review that is a list** | One paragraph per study, each beginning with an author | Reorganise by claim or theme, not by source (Step 4) |
| **Source in the subject position** | "Author A found... Author B found..." throughout | Default to information-prominent citation (Step 5) |
| **The quotation that carries the claim** | Paragraphs ending on a quotation | The claim is yours; quotation supports it (Step 5) |
| **Empty reporting verbs** | Every source "talks about" or "states" | Choose verbs that assert, concede, assume or overlook |
| **Criticality as fault-finding** | Every source is dismissed, none is used | Criticality is assessing what evidence supports (Step 6) |
| **The straw man** | An author's position is simplified, then rejected | State their position at its strongest first (Step 6) |
| **Hedge stacking** | "It may possibly be suggested that it could be" | One hedge, matched to the evidence (Step 7, K3 §4) |
| **Under-hedged speculation** | An interpretation asserted as flatly as a finding | Match register to K3 level (Step 7) |
| **Obscurity mistaken for formality** | Sentences requiring a second reading to parse | Precision, not length; if you cannot follow it, rewrite |
| **Mechanical signposting** | Every section announces and summarises itself | Let the claim signpost; reserve orientation for real joints |
| **Methodology without justification** | The chapter would fit any project | Add the why sentence and a rejected alternative (Step 9) |
| **Discussion repeating results** | No citations, no interpretation, familiar numbers | Enforce the finding-interpretation boundary (Step 9, K2) |
| **Fluent emptiness** | Well-formed prose that makes no claim | Reverse outline; polish only after the argument holds (Step 10) |
| **AI-smoothed prose** | Uniform paragraph shape, confident tone, no specific claim | The student writes; the skill diagnoses (§12.1) |

## 12. AI guardrails

Skill-specific. The universal prohibitions in K4 apply in full and are not repeated.

1. **Academic integrity.** These skills assist a researcher's thinking, structure and rigour. They do not produce work to be submitted as the student's own unaided output. The user must comply with their institution's AI use policy and its declaration requirements, which vary by institution and by assessment. Where an institution prohibits AI assistance for a task, this skill must not be used for it. The skill never writes a passage for submission as though the student wrote it; it interrogates, structures, critiques and teaches. This matters more in this skill than anywhere else in the track, because the writing is itself the thing being assessed. Operationally: diagnose the student's text, name what is missing, show the structure a paragraph needs, ask the question whose answer would be the missing warrant, and let the student write the sentence. Do not supply the sentence. Where a student asks for a paragraph to be rewritten, rewrite one short passage as a demonstration, say explicitly that it is a demonstration and not for submission, and require the student to apply the pattern to the rest in their own words.

2. **Never supply the claim.** A paragraph's claim is the student's position, and it is precisely what the assessment measures. Where a paragraph has evidence and no claim, ask what the student thinks the evidence shows. If they cannot say, the problem is that they have not thought it through yet, and supplying a claim conceals that from both of you and from the marker.

3. **Never supply evidence, examples or supporting detail to complete an argument.** A structurally satisfying paragraph missing its support must not be completed with a plausible finding, a likely figure, or an illustrative source (K4 §2.1, §2.4). Name the gap and ask what evidence the student has. An argument improved by invented support has been made worse.

4. **Never strengthen a claim beyond the evidence in the draft while editing for style.** Editing routinely inflates: hedges get removed for readability, "was associated with" becomes "led to", a qualification is cut for concision. Every such change alters what the work claims and can convert a defensible statement into an unsupportable one (K4 §3.2, K3 §7). Preserve hedges and qualifications, and flag them rather than removing them.

5. **Never impose a single academic style.** Register, use of the first person, permitted directness of criticism, paragraph length and hedging conventions vary substantially between disciplines and traditions. Ask what the field does; where you cannot ask, work from a sample of the field's writing and say what you are inferring from it.

6. **Never smooth prose into uniformity.** Heavily rewritten text acquires a recognisable evenness of rhythm, paragraph shape and confidence that is detectable, that a marker will question, and that erases the student's voice, which is worth preserving both for integrity and because supervisors know their students' writing.

7. **Never present a structural fix as a substitute for reading or analysis.** Where the underlying work is thin, say so. A better-organised version of insufficient material is more visibly insufficient, not less, and telling the student otherwise wastes their remaining time.

8. **Where the draft's argument outruns its evidence, say so at the point it happens** (K2 §2.1, K3 §3.6). The commonest instance is a discussion chapter drawing conclusions the results cannot support, and it is a substantive error, not a writing problem, even though it presents as one.

## 13. Best-practice principles

- **The paragraph is the unit of argument.** Fix paragraphs and chapters improve; fix sentences and nothing changes. This is the single highest-leverage idea in undergraduate academic writing and it is teachable in an hour.
- **Put the claim first and in your own words.** A first sentence that states a position tells the reader what the paragraph is doing, signposts without announcing, and forces you to know what you think before you assemble the evidence.
- **The warrant is where analysis lives.** "And this shows what?" applied to every evidence sentence converts more descriptive writing into analytical writing than any other single intervention.
- **Organise by claim, never by source.** Any structure that gives each source its own paragraph will produce a list, whatever the quality of the individual paragraphs.
- **Read the method sections.** Evaluation requires knowing what a study's design can support, and that information is not in the abstract. This is the practical reason criticality depends on real reading rather than on writing technique.
- **Disagree specifically, and represent the position at its strongest first.** A criticism aimed at a named claim on recognised grounds reads as competence; the same point aimed at a whole paper reads as posturing.
- **Hedge in proportion, once.** Hedging a strong finding is miscommunication, not caution, and stacked hedges signal defensive writing to any experienced reader.
- **Formality is precision, not difficulty.** If you cannot follow your own sentence on a second reading, the sentence has failed, and no marker will assume the fault is theirs.
- **Reverse outline before you edit.** Ten minutes of first sentences tells you more about a chapter than a day of reading it through, because reading through is generous to your own intentions.
- **Cut description to fund argument.** Reporting sentences that summarise a source without using it are the cheapest words in any draft, and cutting them is usually where the words for the missing analysis come from.
- **Each section must earn the next.** If a section could be removed and the following one still work, it is not part of the argument, and length is not the same as development.

## 14. Worked example

Generic fictional scenario, academic, business and management.

**INPUT**

An honours student in management submits a 4,000 word literature review chapter on employee turnover in professional services firms. It cites twenty-two sources, is well referenced, contains no grammatical errors, and has received the feedback: "descriptive, needs critical engagement, and the review does not lead anywhere".

**PROCESS**

*Step 1.* The sentence-type diagnostic is run on two pages. Of forty-one sentences, thirty-four are reporting, five are analysis, and two are positioning, both in the final paragraph of the chapter. The student's response is that they thought summarising accurately was the task, which is the misunderstanding the whole exercise exists to correct.

*Step 3.* The reverse outline is decisive. Twenty-two first sentences are written out. Nineteen begin with an author's name. Read in sequence they form no chain: any two could be swapped without consequence. This is shown to the student rather than described, and it settles the question of whether the chapter argues.

*Step 4.* The chapter is reorganised. Each source is written on a card with what it claims, its population, its measure of turnover, and its design. Sorting the cards produces four positions rather than twenty-two summaries: studies treating turnover as an individual decision, studies treating it as an organisational outcome, studies measuring intention to leave, and studies measuring actual departure. The fourth grouping is the one that matters, and it emerges only from the sort. Most of the reviewed studies measure intention; the few measuring actual departure report weaker and different relationships; and the student's own project measures actual departures. The review now has something to argue, and the gap it will establish is a consequence of the comparison rather than an assertion appended to it.

*Step 2.* One paragraph is rebuilt as a demonstration. Its original form summarised three studies of intention to leave, in three sentences each. Its new form opens with a claim, that intention-based measures may overstate attitudinal influences because they capture a disposition many respondents never act on; cites the three studies as evidence of the pattern; adds the warrant, which is that where studies measured both, the association between attitudes and actual departure is consistently smaller than that between attitudes and intention, so the choice of measure is not neutral; and links forward to what predicts departure specifically, which the next paragraph addresses. The student writes the remaining paragraphs to that pattern, and the demonstration paragraph is marked as such and not carried into the submission.

*The judgement call.* The student wants to write that the intention-based literature is "flawed" and that these studies "cannot tell us anything about actual turnover". This is where criticality tips into overreach, and the correction is substantive rather than stylistic. The studies are not flawed; they measure a construct they define, and several of them say explicitly that intention is a proxy. Disagreeing with a whole literature on a ground its own authors have acknowledged is weak criticism and reads as inexperience. The defensible version is narrower and stronger: intention is a valid construct, the inference from intention to behaviour is an additional step that most of these studies do not test, and the transferability of their findings to actual departure is therefore limited and is an open question. That claim is proportionate, is supported by the comparison already made, and gives the student's project its justification. The hedging is calibrated to match: the pattern in the studies measuring both is stated plainly, and the interpretation of why intention overstates attitudinal effects is offered as the most likely reading with the alternative noted.

*Step 5.* Source integration is revised. Nineteen author-prominent openings drop to five, retained where two authors are compared or one is disputed. Reporting verbs change from a uniform "states" and "discusses" to found, assumed, conceded and did not test. Two paragraphs ending on quotations are rewritten so the student's claim closes them.

*Step 9 and Step 10.* The chapter is re-outlined. The first sentences now read as a chain: turnover is studied through two framings, most work measures intention, the few studies measuring both find the relationship weaker, the inference from intention to behaviour is therefore untested in this setting, and this is what the study addresses. Around 700 words of source summary are cut, which funds the added warrants and keeps the chapter inside its limit.

**OUTPUT**

A restructured chapter organised by position rather than by source, with a claim in the first sentence of each paragraph, warrants connecting evidence to claim, a single specific and proportionate criticism of the dominant measurement approach, and a gap statement the chapter has actually earned. The reference list is unchanged. The improvement is entirely structural, which is the point to make explicitly: the student had done the reading, and the marks were being lost in the organisation of it.

`RESEARCHER REVIEW RECOMMENDED` on the controlling claim and on whether the measurement critique is the right organising argument for this discipline and this supervisor (K5 §2.1).

## 15. Advanced usage

**Card sorting for a stubborn review.** Where a review resists reorganisation, put each source on a card with its claim, population, measure and design, and sort repeatedly on different attributes. The sort that produces the most uneven groups is usually the one that reveals the argument, because an uneven distribution means the field has concentrated somewhere, and a concentration is a finding about the literature.

**Writing the argument before the chapter.** For a long document, write the controlling claim of each chapter and the first sentence of each intended paragraph before drafting anything. This is a reverse outline built forwards, it takes an afternoon, and it prevents the most expensive rewrite in student research, which is discovering after 8,000 words that the chapter does not argue.

**Second-language writers.** Separate the three problems, because they need different work and conflating them wastes effort. Argument structure is language-independent and is often already present. Register is learned by pattern from the field's own writing. Hedging is the highest-risk of the three, since under-hedging reads as overclaiming and over-hedging reads as having nothing to say, and neither is a language error, so proofreading will not touch either.

**Where the standard approach does not fit.** In disciplines using structured formats with fixed section requirements, argument concentrates in the introduction and discussion, and the paragraph unit still applies there while the middle sections are governed by completeness. In interpretive and reflexive traditions, the first person and the researcher's positioning are part of the argument rather than a lapse in formality, and imposing impersonal register would damage the work.

## 16. Skill chain

**Recommended previous skills:**
- **15.02 Academic Literature Search Strategy.** Hands over the source set that a review can argue with, since a review cannot engage with literature it does not have.
- **15.04 Research Proposal Writing.** Hands over the problem, the gap and the alignment matrix, which supply the controlling claims of the early chapters.

**Recommended next skills:**
- None within this track; this skill runs through to submission. Track 15B takes the same argumentative discipline into masters-level work, where theoretical framing and a defended position within a paradigm are additionally required.

**Runs well alongside:**
- **15.03 Referencing and Citation Management**, which supplies verified sources and locators, and which this skill assumes when integrating them into sentences.
- **15.01 Academic Research Topic Selection**, whose problem statement is the first and most important controlling claim the student writes.
- **12.03 Research Report Compilation**, for the contrasting case: where a document must lead with the answer for a reader who is acting rather than assessing.

---
A Yazi Supplied Skill and resource.